Preface:

The Tao (τ) Law, on the surface, is a principle of semiconductor technology, but behind it lies the rule for the second half of localization substitution: the era of single-point breakthroughs is passing, and system capabilities determine the final outcome.

Author | Fang Wensan

Image Source | Internet

From 'Geometric Miniaturization' to 'Temporal Miniaturization'

For more than half a century, the global semiconductor industry has advanced along Moore's Law.

However, today, relying solely on geometric miniaturization to drive chip progress is becoming increasingly difficult.

When transistors shrink to a few nanometers, quantum tunneling effects begin to manifest, power consumption and heat dissipation issues become difficult to solve, and manufacturing costs rise exponentially.

The cost per transistor at advanced nodes no longer continues to decline, with leading chip design budgets exceeding the billion-dollar level. The traditional industrial contract of geometric miniaturization no longer holds.

Advanced processes are like a highway jointly followed by the global industry, where the strongest players often possess the most advanced equipment, processes, and manufacturing resources.

This path remains important today, but its marginal returns are diminishing, entry barriers are rising, and external restrictions prevent Chinese companies from easily catching up along the same path.

Huawei's 'Tao Law' focuses on shifting the main axis of semiconductor evolution from 'geometric miniaturization' to 'temporal miniaturization.'

In other words, technological progress comes not only from smaller transistor sizes but also from shortened signal propagation paths, reduced system communication latency, and improved efficiency in hardware-software-chip collaboration.

Devices, circuits, chips, and systems no longer optimize their respective metrics independently but collectively focus on a single goal: reducing the time constant τ.

With the goal of systematically reducing the time constant τ, core technologies such as 'logic folding' are employed to trade system integration for device miniaturization, continuously compressing signal propagation delays, and continuously improve (continuously improving) chip computing speed, throughput, and overall performance.

Simply put, Moore's Law compresses time by shrinking space, while the Tao Law directly targets time itself as the optimization goal.

Just as building roads on flat ground becomes increasingly difficult, switching to elevated highways, raising speed limits, and optimizing traffic signals can still improve traffic efficiency as long as cars move faster.

It shifts competition from pure process nodes back to system engineering. While transistors remain important, interconnection, circuit layout, packaging, buses, memory, software scheduling, and workload also determine the final experience.

In complex scenarios like AI computing, smart terminals, and server clusters, performance bottlenecks often lie not in a single chip unit but in how data flows, how computing power is organized, how memory is accessed, and how systems collaborate.

This is the industrial implication of the 'Tao Law': When single-point resources are limited, those who can elevate the problem from 'missing a component' to 'reconstructing a system' are more likely to escape passive pursuit.

The Survival Rule of the Second Half of Localization Substitution

The first half of localization substitution focused on supply chain repair, while the second half requires system reconstruction.

China's tech industry has undergone an intensive period of catching up.

Chip design needed to catch up on EDA and IP, manufacturing on equipment and materials, packaging and testing on advanced packaging, complete machines on operating systems and ecosystems, and industrial software on underlying toolchains.

Many companies seized opportunities during this cycle: imported supplies posed risks, domestic supplies entered verification, customers provided windows, and capital offered patience.

In the first half of localization substitution, many companies adopted a single-point breakthrough strategy, concentrating efforts on overcoming a single 'bottleneck' link. This strategy achieved some initial success but also had limitations.

The second half is far more complex, with the real barriers becoming three things: ① Whether a company can participate in defining the customer's next-generation architecture. ② Whether it can embed its products into the customer's long-term technical roadmap. ③ Whether it can transform the substitution relationship into a symbiotic one through continuous iteration.

At this point, localization substitution shifts from a procurement behavior to an engineering behavior, from a supply chain security issue to an industrial efficiency issue.

System innovation becomes the core competitiveness in the second half of localization substitution. Through system-level innovation, deficiencies in individual device performance can be compensated for, achieving a leap in overall performance.

During supply chain reconstruction, domestic suppliers were nurtured through investment, technical support, and order support, fostering a 'symbiotic and mutually prosperous' ecological relationship. This ecological collaboration model will accelerate the process of localization substitution.

The value of the 'Tao Law' lies in its attempt to escape the anxiety of merely following advanced process routes. It does not deny the importance of process progress but places system efficiency on an equal footing.

Against the backdrop of shortcomings in advanced lithography, core manufacturing equipment, high-end materials, and full EDA workflows, simple pursuit would keep companies under long-term pressure.

A more realistic path is to maximize architecture, packaging, interconnection, software, algorithms, and scenario understanding under available manufacturing conditions, making system performance approach or even locally surpass user needs.

This is not about bypassing hard technology or lowering standards. On the contrary, it requires stronger engineering comprehensive ability (comprehensive engineering capabilities).

Localization Substitution Must Create New Performance Curves

In recent years, China's semiconductor industry chain has seen the emergence of numerous niche champions, with significant progress in RF, power, analog, MCU, sensors, equipment components, packaging materials, and partial EDA tools.

However, from an industrial competition perspective, niche champions do not equate to system victory. Semiconductor is a highly coupled industry, and the more advanced it goes, the more it tests collaboration.

Materials must match equipment, equipment must match processes, processes must match design, design must match packaging, packaging must match systems, and systems must match software ecosystems.

The difficulty of the second half lies here: domestic companies need to move from point-like breakthroughs to chain-like collaboration.

Progress by a single etching equipment company is insufficient for advanced manufacturing breakthroughs; partial tool progress by an EDA company is insufficient to support a complete design closed loop (closed loop).

Only when equipment, materials, EDA, manufacturing, packaging, software, and applications iterate together can the industry develop true resilience.

In the past, we always asked: How many generations behind is advanced process technology? When will lithography machines break through? When will EDA be fully substituted? When will domestic chips fully catch up?

Now, the questions have shifted: Can system-level innovation reshape efficiency under existing constraints? Can hardware, software, architecture, interconnection, and applications be optimized together? Can China's vast application scenarios feed back into underlying technologies?

The rigor of the second half lies in the fact that the easily plucked fruits have already been picked. Remaining challenges, such as high-end GPUs, advanced lithography processes, aerospace-engine-level industrial software, and chemical process simulation systems, require investment cycles spanning decades and long-term accumulation of basic research.

However, most companies cannot replicate their hundred-billion-dollar R&D scale, which forces another important rule: hierarchical relay.

Leading companies undertake the foundation and general-purpose platforms, while SMEs specialize in niche areas to the extreme, forming a layered and symbiotic industrial chain. The value of specialized and innovative companies will gradually outweigh platform-centric approaches, ultimately delivering new performance curves.

Examples include equipment parameters better suited to local processes, software architectures closer to Chinese customer scenarios, interconnection schemes more suitable for AI servers, power device combinations more fitting for new energy vehicles, and stability and maintenance costs more appropriate for industrial scenarios.

Substitution is not about replacing an imported part with a domestic one; the highest stage of substitution is bringing the customer's original technical roadmap onto a new efficiency curve.

Conclusion:

Localization substitution has reached a point where it can no longer rely solely on sentiment and windows of opportunity. True substitution must grow from the depths of the supply chain, developing into new system capabilities and new industrial rules.

Transform single-point products into system value, short-term orders into long-term roadmaps, and substitution opportunities into definition capabilities.

The 'Tao Law' is like planting a new coordinate in the semiconductor industry, reminding tech companies that true breakthroughs are not just about catching up with others' paths but also about having the ability to forge their own under constraints.

Partial References: 36Kr: 'The Forced 'Tao (τ) Law': A Hidden Experiment Behind Huawei's 381 Chips,' People's Review: 'Semiconductors Embrace the 'Tao (τ) Law': China's Definition Will Rewrite the World,' TMTPost APP: 'τ Law: China's First Attempt to Define the Future of the Semiconductor Industry,' Guancha.cn: 'Dialogue with Wang Bo: What Are the Limits of Huawei's 'Tao Law?',' Nanfang Plus: 'From 'Spreading Out' to 'Folding': An Integrated Circuit Expert Interprets Huawei's Tao Law'